22 Jul 2025

AI News Digest

🤖 AI-curated 6 stories

Today's Summary

Hey, so OutSystems just dropped Agent Workbench, a new tool to help enterprises integrate AI agents seamlessly into their workflows, streamlining processes and tackling that annoying siloed data problem. Meanwhile, ProductNow snagged $6 million to build an AI operating system for product teams, aiming to centralize planning and reduce manual tasks—a sign that AI’s not just for engineers anymore. And on the creative side of things, researchers launched IRIS to boost scientific innovation with AI, showing that even hypothesis brainstorming is getting an AI upgrade.

Stories

OutSystems Debuts Agent Workbench to Help Enterprises Build and Scale AI Agents

OutSystems has launched Agent Workbench, a tool designed to help enterprises develop intelligent AI agents to improve human-AI collaboration. This offering enables organizations to streamline workflows by integrating agentic AI into existing systems seamlessly, supporting major providers like Azure, OpenAI, and AWS Bedrock, while offering centralized AI governance and data connectivity. OutSystems aims to shift enterprise AI from experimental to operational by addressing challenges of siloed data and fragmented tools.

ProductNow Raises $6M to Build AI Operating System for Product Teams

ProductNow has secured $6 million to develop an AI operating system for product, program, and strategy teams. Founded in 2025, it aims to replace fragmented tools with a centralized AI system, automating planning, task assignment, and reporting by integrating AI agents with existing tools. This approach targets roles beyond just engineering. Led by Sierra Ventures, this seed round supports ProductNow's vision of transforming strategic goals into actionable roadmaps, focusing on seamless execution from business intent to impact, with continuous progress monitoring and reduced manual workflows across various functions.

IRIS: Interactive Research Ideation System Enhances Scientific Discovery

Researchers have introduced IRIS, an open-source platform designed to assist scientists in generating novel hypotheses. By integrating large language models with features like adaptive compute expansion and fine-grained feedback, IRIS aims to accelerate the research ideation process. A user study across various disciplines validated its effectiveness in enhancing scientific creativity.

Bridging Econometrics and AI: Reinforcement Learning Meets GARCH Models for Risk Estimation

A new study proposes a hybrid framework combining GARCH volatility models with deep reinforcement learning to improve Value-at-Risk (VaR) estimation in volatile financial markets. By integrating market forecasting using Double Deep Q-Networks, the approach dynamically adjusts risk-level forecasts, demonstrating significant improvements in accuracy and regulatory compliance.

Google's 'Try On' AI Feature Enhances Online Shopping Experience

Google has launched its AI-powered 'Try On' feature in Search for U.S. users, allowing them to virtually try on clothes by uploading full-length photos. Integrated with Google's Shopping Graph, this tool provides realistic images of how clothing items would look on the user, enhancing the online shopping experience.

Microsoft Introduces AI-Powered Features in Windows 11 Update

Microsoft's upcoming Windows 11 update in August 2025 will include several AI-powered features such as 'Copilot Vision' for real-time desktop interaction, an AI Agent in Settings for natural language commands, and 'Relight in Photos' for AI-enhanced photo editing. These enhancements aim to improve user productivity and experience.